Purpose: Bulk drug substance is often stored frozen to control product inventory and extend shelf life relative to refrigerated storage. Characterizing the freezing of biologics at full scale can be expensive since large volumes of drug substance are required. An understanding of both heat and mass transfer is necessary to develop a representative scale-down model to describe bulk-scale freezing of biologics. Here we describe the development and validation of computational models to predict freeze duration and macroscopic cryoconcentration of bulk drug substance freezing in a liquid nitrogen (LN2) blast freezer. Methods: Computational fluid dynamics software was used to predict freeze durations and concentration gradients on freezing of bulk protein solutions in a LN2 blast freezer. In order to assess spatial variability of air temperature and velocity within the LN2 blast freezer a computational model was developed containing eight disposable containers (“freezer model”). The boundary conditions defined in the freezer model were used to develop a second model, which was used to estimate freeze duration heterogeneity throughout the freezer (“container model”). Convective flow due to thermally-driven density differences was neglected for these simulations to improve the speed and stability of the computational model for determination of freeze durations. Lastly, model refinements including solute‑density driven convection were reintroduced into the container model to simulate protein cryoconcentration that occurs during freezing. Models were validated using experimental data. Results: Heat transfer conditions between the container and the freezer (e.g. heat transfer coefficients and gas temperature) obtained from the LN2 freezer model were in reasonable agreement with experimental measurements and are suited for use for freeze duration prediction in a computational model of a single container. Predictions of freeze duration in a 6‑L container were in reasonable agreement with experimental values, while predictions made for the 12-L container were approximately 15% underpredicted. Inclusion of solute-density driven convection improved the prediction of cryoconcentration enhancement factor and vertical stratification on freezing. By understanding the heat transfer and solution properties of drug substance during bulk scale freezing, a representative scale down model can be developed and used in place of bulk scale studies.

Computational modeling of bulk scale freezing to predict freeze duration and cryoconcentration of biologics in a liquid nitrogen blast freezer

249th National Meeting (2015)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.